A comparison of auditory and blind separation techniques for speech segregation.pdf

IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 9, NO. 3, MARCH 2001 189 A Comparison of Auditory and Blind Separation Techniques for Speech Segregation André J. W. van der Kouwe, Member, IEEE, DeLiang Wang, Member, IEEE, and Guy J. Brown Abstract—A fundamental problem in auditory and speech pro- cessing is the segregation of speech from concurrent sounds. This problem has been a focus of study in computational auditory scene analysis (CASA), and it has also been recently investigated from the perspective of blind source separation. Using a standard corpus of voiced speech mixed with interfering sounds, we report a compar- ison between CASA and blind source separation techniques, which have been developed independently. Our comparison reveals that they perform well under very different conditions. A number of conclusions are drawn with respect to their relative strengths and weaknesses in speech segregation applications as well as in mod- eling auditory function. Index Terms—Auditory scene analysis, blind source separation, computational auditory scene analysis (CASA), oscillatory corre- lation, speech segregation. I. INTRODUCTION H UMAN listeners exhibit a remarkable ability to segregate the voice of a single speaker from a mixture of other in- truding sounds. This phenomenon may be regarded as one as- pect of a more general process of auditory organization, which is able to untangle an acoustic mixture in order to retrieve a percep- tual description of each constituent sound source. The term au- ditory scene analysis (ASA) has been introduced to describe this process [4]. Conceptually, ASA may be regarded as having two stages. In the first stage, the acoustic mixture is decomposed into sensory elements (“segments”). The second stage (“grouping”) then combines segments that are likely to have originated from the same sound source. Recently, attempts to develop computational systems that mimic ASA have led to the emergence of a new field, known as computation